Cardiovascular Magnetic Resonance Imaging for the Investigation of Ischaemic Heart Disease

Introduction: Coronary artery disease (CAD) remains the number one cause of mortality worldwide; improving diagnosis and treatment is a priority. Multi- parametric cardiovascular magnetic resonance (CMR) offers quantitative assessment of the cardiovascular system with a variety of techniques allowing assessment of anatomy, function, myocardial composition and perfusion during a single scan. Aims: To assess 1.) diagnostic accuracy of visual and quantitative perfusion CMR to single-photon emission computed tomography (MPS-SPECT) in patients with left main stem CAD. 2.) the hypothesis that patients with ischaemic (ICM) and non-ischaemic cardiomyopathy (NICM) have different torsion and strain parameters 3.) development and validation of a contemporary multivariable risk model of CAD from a large population undergoing X-ray angiography. 4.) a rapid 3D mDIXON pulse sequence for image quality and quantitation of MI. 5.) T1 rho prepared (T1ρ) dark blood sequence and compare to blood nulled PSIR (BN) and standard myocardium nulled PSIR (MN) for detection and quantification of scar. Methods: Patients were recruited between 2008 and 2017. Patients in chapters 3,4,6,7 underwent multi-parametric CMR including late gadolinium enhancement (LGE) imaging at 1.5 or 3.0T. Patients in chapter 5 underwent angiography. Results: 1.) CMR demonstrated significantly higher area under the curve for detection of LMS or equivalent disease over MPS-SPECT(P=0.0001). 2.) Despite no difference in LV dimensions, EF and strain between ICM and NICM, NICM patients had significantly lower LV twist(P=0.023) and torsion(P=0.017) compared to ICM. 3.) The developed model discriminated well and was well-calibrated. Diamond and Forrester and Duke scores substantially over-predicted CAD risk, whilst CAD Consortium risk models slightly under-estimated risk. 4.) Image quality was comparable between 3D and 2D LGE(P=0.162). Time for 3D image acquisition was only 5% of the time required for a standard 2D acquisition. 5.) CNRscar-blood was significantly increased for BN and T1ρ compared to MN LGE. BN LGE demonstrated significantly higher reader confidence scores.